Direct Sequence Spread Spectrum (DSSS) has been widely used in military and civilian communications due to its strong resistance to various common interferences, high security, and ease of implementation. It has been widely used in Code Division Multiple Access (CDMA) system. However, in non-cooperative communication scenarios, detecting DSSS signals, estimating DSSS signal parameters, and even intercepting information are all issues that need to be considered. In DSSS, correctly identifying the spread spectrum sequence used is an important prerequisite for correcting despreading. To address the problem of low success rate of spread code identification for low signal-to-noise ratio DSSS signals, this paper combines the Third-order Correlation Function (TCF) of m-sequences and its peak characteristics to identify the pseudo-code period of DSSS signals as prior information through power spectrum secondary processing on the premise of denoising preprocessing. The problem of spread code identification is transferred into a peak detection classification problem. The peak identification and classification is then studied.

This paper proposes a method of using Sparrow Search Algorithm (SSA) to optimize Extreme Gradient Boosting (XGBOOST) for third-order correlation peak classification of direct spread signals to improve the accuracy of m-sequence classification and identification.

By comparing conventional peak detection and decision tree classification methods at different signal-to-noise ratios and comparing the classification accuracy of different sequence periods, the simulation results show that the spread code identification and classification method optimized by SSA with XGBOOST after preprocessing has a higher classification and identification success rate than conventional machine learning and peak detection methods. Its performance gradually improves at high sequence periods.

This method can more accurately identify and classify m-sequence spread codes under low signal-to-noise ratio conditions.